Active Short Selling by Hedge Funds - European Corporate ...

Page created by Harold Luna
 
CONTINUE READING
Active Short Selling by Hedge Funds

  Finance Working Paper N° 609/2019   Ian Appel
                          May 2019    Boston College

                                      Jordan Bulka
                                      Boston College

                                      Vyacheslav Fos
                                      Boston College and ECGI

                                      © Ian Appel, Jordan Bulka and Vyacheslav Fos
                                      2019. All rights reserved. Short sections of text, not
                                      to exceed two paragraphs, may be quoted without
                                      explicit permission provided that full credit, includ-
                                      ing © notice, is given to the source.

                                      This paper can be downloaded without charge from:
                                      http://ssrn.com/abstract_id=3242516

                                      www.ecgi.global/content/working-papers
ECGI Working Paper Series in Finance

            Active Short Selling by Hedge Funds

                                      Working Paper N° 609/2019
                                                 May 2019

                                             Ian Appel
                                           Jordan Bulka
                                          Vyacheslav Fos

We have benefitted from comments by Alon Brav, Cliff Holderness, Robert Jackson, Wei Jiang, Peter Kelly,
and Jeff Pontiff and seminar/conference participants at Bentley University, The Chinese University of Hong
Kong, Fanhai International School of Finance, UC Berkeley Law, Economics, and Business workshop, the
SEC, Yale Junior Finance Conference, NYU Law & Finance seminar, the RCFS/RAPS Conference at Baha
Mar, and the 2019 American Law and Economics Association Meetings. We also thank Sourabh Banthia, Troy
Heidenberg, Ryan Leary, Vinh Nguyen, Spencer Olson, Fisher Pressman, and Jonah Schumer for excellent
research assistance. This paper previously circulated under the title “Public Short Selling by Activist Hedge
Funds.”

© Ian Appel, Jordan Bulka and Vyacheslav Fos 2019. All rights reserved. Short sections of text,
not to exceed two paragraphs, may be quoted without explicit permission provided that full credit,
including © notice, is given to the source.
Abstract
Short selling campaigns by hedge funds have become increasingly common in
the last decade. Using a hand-collected sample of 252 campaigns, we document
abnormal returns for targets of approximately -7% around the announcement
date. Firm stakeholders, including the media, plaintiffs’ attorneys, and other short
sellers, play an important role in campaigns. Changes in aggregate short interest
do not drive the effects on firm value and stakeholder behavior. Campaigns are
primarily undertaken by activist hedge funds. Evidence suggests disclosure costs
and information are important channels through which activism technology affects
short selling.

Keywords: Activist Hedge Funds, Short Selling

JEL Classifications: G23, G14

            Ian Appel*
            Assistant Professor of Finance
            Boston College, Carroll School of Management
            140 Commonwealth Avenue
            Chestnut Hill, MA 02467-3809, United States
            phone: +1 617 552 1459
            e-mail: ian.appel@bc.edu

            Jordan Bulka
            Researcher
            Boston College, Carroll School of Management
            140 Commonwealth Avenue
            Chestnut Hill, MA 02467-3809, United States
            e-mail: jordan.bulka@bc.edu

            Vyacheslav Fos
            Associate Professor of Finance
            Boston College, Carroll School of Management
            140 Commonwealth Avenue
            Chestnut Hill, MA 02467, United States
            phone: +1 617 552 1536
            e-mail: fos@bc.edu

            *Corresponding Author
Active Short Selling by Hedge FundsI

                                       Ian Appel
                    Carroll School of Management, Boston College
                                   ian.appel@bc.edu

                                     Jordan Bulka
                    Carroll School of Management, Boston College
                                jordan.bulka@bc.edu

                                    Vyacheslav Fos
                    Carroll School of Management, Boston College
                                vyacheslav.fos@bc.edu

                                   First Version: March 2018
                                    This Version: May 2019

  I
     We have benefited from comments by Alon Brav, Cliff Holderness, Robert Jackson, Wei Jiang, Peter
Kelly, and Jeff Pontiff and seminar/conference participants at Bentley University, The Chinese University of
Hong Kong, Fanhai International School of Finance, UC Berkeley Law, Economics, and Business workshop,
the SEC, Yale Junior Finance Conference, NYU Law & Finance seminar, the RCFS/RAPS Conference
at Baha Mar, and the 2019 American Law and Economics Association Meetings. We also thank Sourabh
Banthia, Troy Heidenberg, Ryan Leary, Vinh Nguyen, Spencer Olson, Fisher Pressman, and Jonah Schumer
for excellent research assistance. This paper previously circulated under the title “Public Short Selling by
Activist Hedge Funds.”
Active Short Selling by Hedge Funds

Abstract

Short selling campaigns by hedge funds have become increasingly common in the last
decade. Using a hand-collected sample of 252 campaigns, we document abnormal returns for
targets of approximately -7% around the announcement date. Firm stakeholders, including
the media, plaintiffs’ attorneys, and other short sellers, play an important role in campaigns.
Changes in aggregate short interest do not drive the effects on firm value and stakeholder
behavior. Campaigns are primarily undertaken by activist hedge funds. Evidence suggests
disclosure costs and information are important channels through which activism technology
affects short selling.
Short sellers have been cast as villains throughout history. Following the crash of 1929,
for example, the U.S. Senate released the names of large short sellers in an attempt to
brand them as “unpatriotic” (Jones, 2012). More recently, the Attorney General of New
York likened short sellers to “looters after a hurricane” during the financial crisis. Perhaps
because of this sentiment, investors are often reluctant to disclose short positions. Yet,
recent years have seen a new phenomenon: high-profile short selling campaigns by hedge
funds. David Einhorn’s short of Allied Capital provides an illustrative example. In May of
2002, Einhorn announced a short position in Allied at an investment conference, arguing
the firm engaged in questionable accounting practices. Allied’s stock dropped over 10% the
following day, and by the next month its short interest increased six-fold. The SEC later
launched an investigation into Allied that “zero[ed] in on many of the criticisms made by
short sellers.”1
       In this paper, we undertake a comprehensive analysis of public short selling campaigns
by hedge funds. We refer to these campaigns as “active” short selling because hedge funds
take costly actions (e.g., public disclosure) to enhance the flow of negative information
into prices. The costs associated with such campaigns can be substantial. First, disclosure
invites regulatory scrutiny. For example, authorities investigated Bill Ackman for market
manipulation related to his short position in Herbalife.2 Second, firms can take various
actions to impede short sellers. Such actions include encouraging investors to withdraw
shares from the lending market, limiting access to conference calls, and litigation (Lamont,
2012). Third, to the extent that short sellers are informed, disclosures reduce information
asymmetries and may limit future trading opportunities.3
       A priori, the extent to which hedge funds engage in active short selling is unclear. On
one hand, hedge funds may be likely candidates to undertake such campaigns because,

   1
     See “SEC Is Investigating Allied Capital,” The Wall Street Journal (6/24/2004)
   2
     See “Prosecutors Interview People Tied to Ackman in Probe of Potential Herbalife Manipulation,” The
Wall Street Journal 3/12/2015
   3
     Consistent with short selling campaigns being costly, Jones et al. (2016) and Jank et al. (2016) find
that mandatory disclosure of short positions leads to a decrease in short selling.

                                                    1
in contrast to other types of investors (e.g., mutual funds), they face few investment
restrictions related to short selling. Moreover, many hedge funds engage in shareholder
activism and are accustomed to taking hostile actions against firms. On the other hand,
hedge funds, in aggregate, have trillions of dollars under management and use contractual
provisions that mitigate risks associated with short selling (e.g., gate provisions). Thus,
limits to arbitrage that motivate disclosure of short positions by small investors may not
bind for hedge funds (Ljungqvist and Qian, 2016).
     We manually construct a database of active short selling campaigns by hedge funds
from media reports. We identify 252 campaigns over a twenty year period (1996–2015).
Consistent with anecdotal evidence, the prevalence of active short selling campaigns has
increased considerably in recent years. Prior to 2008, the average number of campaigns
was fewer than 10 per year, but starting in 2008 the average number of campaigns
approximately tripled, peaking with 45 in 2015 (see Figure 1). The campaigns feature
a wide array of allegations; the most common are general overvaluation, problems with the
industry/competitors, defective product/business model, and fraud.
     We next turn to the central question we ask in this paper: What economic factors
influence the decisions of hedge funds to engage in active short selling?
     To shed light on this question, we first analyze the benefits of active short selling
campaigns. Targets experience cumulative abnormal returns (CARs) of about -3% in a [t–
10, t+10] window around campaign announcements. This negative abnormal performance
is not short-lived; CARs decrease to less than -7% for the [t–10, t+100] window. The
magnitudes of the effects are considerably larger than CARs associated with large changes in
short interest, suggesting that active short selling has an incremental impact on targets. To
mitigate concerns related to confounding events (e.g., negative news), we find similar results
for campaigns announced at investment conferences, the timings of which are predetermined
and likely orthogonal to other announcements.
     We also find that active short selling is associated with changes in the behavior of
other stakeholders, potentially to the benefit of hedge funds initiating campaigns. First,

                                              2
campaigns are associated with approximately a 10% increase in aggregate short interest,
perhaps reflecting strategic complementarities between speculators (Goldstein et al., 2013).
Second, active short selling is associated with an increase in media coverage, particularly
for articles with a negative tone. Finally, campaigns are associated with an increase in
the likelihood of litigation (e.g., shareholder class actions) against firms. Similar to the
valuation results, the effects on media coverage and litigation are not driven by changes in
aggregate short interest, suggesting that the public engagement aspect of campaigns has
an incremental impact on stakeholder behavior.
       Next, we examine the characteristics of hedge funds that engage in active short
selling. We find that campaigns are predominately undertaken by activist hedge funds.4
Specifically, activists initiate 208 out of 254 campaigns in our sample (82%). At the fund-
quarter level, the likelihood of engaging in active short selling is 0.56% for activist hedge
funds, relative to the unconditional likelihood of 0.16%. Other fund characteristics (e.g.,
portfolio liquidity, concentration, and turnover) have relatively little explanatory power for
undertaking an active short selling campaign.
       Our analysis of fund characteristics suggests that the distinction between activist and
non-activist hedge funds is important for understanding what drives active short selling.
Evidence points to two mechanisms that potentially explain this difference in behavior:
costs of public engagement and information.
       Perhaps the most salient costs associated with active short selling are those stemming
from confrontations with the target, long shareholders, or regulators. Activist hedge funds
may be uniquely suited to bear such costs. A fundamental difference between activist and
non-activist investors is that activists often engage in confrontational long campaigns in
order to influence the corporate policies of targets (Brav et al., 2008). Because activists
already have reputations as confrontational investors, the incremental costs associated with
active short selling campaigns may be lower. Consistent with this idea, we find that active

   4
    We define a hedge fund as an activist if it has filed at least one Schedule 13D (Brav et al., 2008). In
our dataset, 240 out of 1,199 hedge funds are activists.

                                                    3
short selling is positively associated with activism experience (i.e., the number of previous
long activism campaigns) as well as with the use of hostile tactics (e.g., proxy fights) in
long activism campaigns.
     We also consider the role of information in active short selling campaigns. While
many equity investors base investment decisions on quantitative signals, the investment
process for activists is fundamentally different. Specifically, activists identify problems
at potential targets (e.g., ineffective management). If a problem is both priced and can
be corrected (e.g., by changing the management team), the fund may initiate an activism
campaign. If the problem is not priced, the economic incentives for long activism are absent
as prices are likely to drop if an activist reveals the problem. There are, however, strong
incentives to engage in active short selling in this case. We find two pieces of evidence
consistent with this mechanism. First, two-thirds of campaigns feature specific allegations
of problems for firms (e.g., fraud) rather than a general claim of overvaluation. Second,
campaigns with specific allegations are associated with significantly lower announcement
returns, suggesting that such information is conducive to active short selling campaigns.
     We find less evidence supporting alternative mechanisms. First, benefits of active
short selling (e.g., as measured by CARs) are similar for activist and non-activist hedge
funds. Second, limits to arbitrage are unlikely to explain the differences in behavior
between these types of investors. Third, we do not find reputational effects stemming
from campaigns. Finally, active short selling is not used to hedge existing long positions.
     While activist hedge funds are prominent in our sample, it is important to note key
differences between shareholder activism and short selling campaigns. In a long activism
campaign, an activist identifies a company whose future free cash flows can be increased if a
particular action is taken. The activist engages with the target to achieve this action, often
using (or threatening to use) decision rights inherent to stock ownership (e.g., voting or
board representation). Short sellers, however, do not have the rights associated with equity
ownership. Rather than trying to convince management to take a particular action, most
active short selling campaigns instead seek to enhance the flow of negative information into

                                              4
prices. We therefore do not generally consider such campaigns a form of activism.
       Our paper is related to multiple strands of literature. First, we contribute to the
literature on short selling. A number of papers show short interest is associated with
weaker future performance at both the market and individual stock levels.5 A related
strand of literature studies informational advantages of short sellers, including the detection
of financial misconduct (Karpoff and Lou, 2010), anticipation of earnings and analyst
downgrades (e.g., Christophe et al., 2004, 2010), and the ability to process public news
releases (Engelberg et al., 2012). In contrast to this literature, we examine active short
selling campaigns in which positions are voluntarily disclosed. While such campaigns have
been rare historically, they have become increasingly common in recent years.
       Second, our paper is related to recent work on public short selling by different market
participants. Ljungqvist and Qian (2016) show that negative research reports help small
investors to overcome limits to arbitrage. Zuckerman (2011) studies the use of public
recommendations of long and short positions by hedge funds as a coordinating mechanism.
Zhao (2019) studies the relation between accounting opacity and public short selling by
financial institutions, research firms, and individual investors. In contrast to this other
work, we highlight the presence of economic synergies between short selling and access
to activism technology. These synergies not only affect the manner in which hedge funds
engage in short selling, but ultimately have implications for the effects of short selling on
targeted firms.
       Third, our paper is related to the literature on hedge fund activism. This literature
broadly finds that long campaigns by activists are associated with positive stock price
reactions and improved operating performance (e.g., Brav et al., 2008; Becht et al., 2008;
Klein and Zur, 2009; Bebchuk et al., 2015; Clifford, 2008).6 While this literature focuses on

   5
     For example, see Seneca (1967), Senchack and Starks (1993), Desai et al. (2002), and Rapach et al.
(2016). Previous work has also analyzed alternative measures of aggregate or institutional short selling,
including required disclosure in European markets (e.g., Jones et al., 2016), costs associated with shorting
(e.g., Cohen et al., 2007; Jones and Lamont, 2002), and institutional short sale orders (e.g., Boehmer et al.,
2008). See Reed (2013) for a comprehensive review of this literature.
   6
     Other papers examine the effect of activists on various outcomes, including innovative activities (Brav

                                                      5
the long positions of activist hedge funds, we provide some of the first evidence on public
disclosure of short positions by this class of investors.

1. Data

       Our sample consists of active short selling campaigns by hedge funds between 1996
and 2015. Because U.S. securities laws do not mandate disclosure of short positions, we
collect information on campaigns from public sources. To construct the sample, we begin
with a list of more than 1,200 hedge funds from Agarwal, Fos, and Jiang (2013). This
list contains hedge funds that have filed a Form 13F with the Securities and Exchange
Commission (SEC).
       We use Factiva to create a database of publicly disclosed short positions by hedge
funds. Specifically, we conduct searches of the form “Hedge Fund Name” and (“short
position” or “short selling”), as well as other variants of these phrases across the
newspapers, newswires, blogs, and television transcripts covered by Factiva. In some cases
we conduct internet searches to find additional information. For each campaign, we collect
the date of the announcement, the revelation method (e.g., investment conference, media
interview, release of a white paper, etc.), and the primary allegations made by the hedge
fund (e.g., general overvaluation, fraud, threats from competitors, etc.). We match target
firms to their respective identifiers by searching the CRSP/Compustat merged database.
The final sample consists of 252 short selling campaigns by 51 hedge funds from 1996–2015.
       We merge data from several other sources to our sample. Financial accounting
and stock return information are from the Compustat annual files and CRSP daily files,
respectively. Short interest and fails-to-deliver data are from Compustat and the SEC,
respectively. Institutional ownership data are from Thomson Reuters and Form 13F filings.7

et al., 2018), productivity and asset allocation (Brav et al., 2015), and takeover offers (Boyson et al., 2017;
Greenwood and Schor, 2009). See Brav et al. (2010), Denes et al. (2017), and Gillan and Starks (2007) for
comprehensive reviews of this literature.
   7
     Form 13F, which discloses quarter-end holdings of an institution, is mandatory for all institutions that
exercise investment discretion over $100 million of assets in equity and other publicly traded securities.
Institutions are required to disclose all securities that appear on the official list of “Section 13(f) Securities,”

                                                         6
Analyst coverage is from IBES, and litigation data are from Audit Analytics. Measures of
the quantity and tone of media coverage are from RavenPack News Analytics. All variables
used in our analysis are defined in Table A1.

2. Campaign and Target Characteristics

      In this section, we study the prevalence of active short selling campaigns over time,
as well as their allegations and disclosure methods. We also examine the characteristics of
targets that predict campaigns.

2.1. Campaign Characteristics

      Figure 1 shows the time series distribution of active short selling campaigns. Prior to
2008, the number of campaigns averaged approximately 5 per year. In later years, however,
the number of campaigns increased substantially, averaging 25 per year from 2008–2015.
While some of campaigns launched in 2007–2008 were in direct response to the financial
crisis (e.g., Bill Ackman’s campaign against Lehman Brothers), the number of short selling
campaigns by hedge funds has remained elevated in the post-crisis period.8 Indeed, 2015
saw more short selling campaigns by hedge funds than any other year in our sample. Panel
B shows the number of hedge funds that have launched short selling campaigns during the
sample. The time series pattern is similar to panel A, though the magnitudes are lower
indicating that some hedge funds launch multiple campaigns in the same calendar year.

                                         [Insert Figure 1 here]

      We next examine the allegations of active short selling campaigns. For each campaign
in our sample, we classify the allegations into six categories: financials/capital structure,

including almost all publicly traded equity, some preferred stocks, bonds with convertible features, warrant,
and exchange-traded call and put options. The Thomson Reuters database contains only holdings of equity.
We thank Baozhong Yang for providing data on Form 13F long positions in exchange-traded call and put
options.
   8
     There were two major rule changes related to short selling during the crisis. In September of 2008,
the SEC banned most short selling in nearly 1,000 financial stocks and “abusive” naked short selling in all
stocks (Boehmer et al., 2013). The short selling ban was lifted the following month; restrictions on naked
short selling were made permanent in 2009.

                                                     7
industry/competitors, general overvaluation, fraud/accounting, product/business model,
and management/insider selling. The classifications are not mutually exclusive, so some
campaigns include multiple types of allegations. Panel A of Table 1 reports the distribution
of short campaigns across the types of allegations.
     Approximately 37% of the campaigns allege that the target is generally overvalued
but do not make a specific allegation. The remainder of the campaigns make specific
allegations regarding the target firm. The most common allegations relate to a target’s
industry/competitors (25.0%), product/business model (16.3%), or fraud/accounting
problems (13.5%). Other allegations include issues related to financials/capital structure
(12.1%) and issues with management or insider selling (4.4%).

                                    [Insert Table 1 here]

     Panel B reports the distribution of disclosure methods for active short selling
campaigns.    We classify disclosure methods into four categories: letters to investors,
newspapers/television, investment conferences, and white papers/other. The most common
form of disclosure, in the newspaper or on television, accounts for about 54% of active short
selling campaigns. Another 22.6% of positions are announced at investor conferences.
Approximately 13% of campaigns are disclosed in a letter to the fund’s investors. The
remaining 10.3% are disclosed through a white paper or other methods.

2.2. Target Selection

     Table 2 reports characteristics of active short selling targets as well as the difference
with matched firms. The firm characteristics and matching methodology used in this table
follow Brav et al. (2008). Targets differ from the matched sample along a number of
dimensions. The starkest difference is in terms of size: on average the market cap for
targets is close to $11 billion larger than matched firms. One potential explanation for
this finding is that we construct our sample based on media coverage of campaigns. The
media, of course, is more likely to cover large firms. We match on firm size (as measured
by market cap) in the remainder of our analysis to account for this difference.

                                             8
Targets tend to have higher Q, ROA, and previous stock returns than matched firms,
indicating that they are firms with strong past performance. We also find evidence that
targets, on average, have higher institutional ownership, more analyst coverage, and higher
leverage than matched firms. We do not find evidence of differences in cash holdings or
payouts between the targets of active short selling campaigns and matched firms. However,
targets are more liquid (as measured by Amihud (2002) illiquidity measure).

                                    [Insert Table 2 here]

     Table 3 analyzes which variables predict active short selling campaigns using probit
model (column 1) and OLS models (columns 2–4). The dependent variable is an indicator
equal to one if a firm is targeted by a campaign in year t. The sample consists of all
Compustat firm-year observations from 1996-2015. Similar to Table 2, we find robust
evidence that the likelihood of targeting is positively associated with firm size across
the different specifications. Targeting is also positively associated with leverage (as in
Table 2), but negatively associated with institutional ownership (in contrast to Table 2).
Short interest also predicts campaigns. Specifically, the estimates in column 4 indicate
that a 5 percentage point increase in short interest is associated with a 0.2% increase in
the likelihood of targeting. This relation is economically significant relative to the 0.16%
unconditional likelihood of a campaign.

                                    [Insert Table 3 here]

3. Shareholder Wealth Effects

     We next turn attention to the potential benefits of active short selling. One such
benefit is to enhance the flow of negative information into prices. To explore this idea, we
examine the cumulative abnormal returns (CARs) associated with the announcement of
active short selling campaigns. We also compare the CARs to those from large changes in
short interest to shed light on the incremental effects of active short selling.

                                              9
Table 4 reports the results. The dependent variables in this table are CARs for the
[t–10, t+10] period (columns 1–3) and [t–10, t+100] period (columns 4–6) around the
announcement of an active short selling campaign. We calculate CARs using a one, three,
and five-factor model for each interval. Panel A reports CARs for active short selling
campaigns. We find broad evidence that campaigns are associated with negative abnormal
returns. Specifically, CARs in the [t–10, t+10] period range from -3.4% to -3.6%, while
those for [t–10, t+100] range from -6.0% to -7.9%. In Panel B, we conduct similar tests
for large changes in short interest (i.e., more than 5 percentage points). The economic
magnitudes of these effects are considerably smaller (and in some cases positive).9 Panel C
reports the difference between CARs for active short selling campaigns and large changes
in short interest. The differences are economically large and statistically significant at
conventional levels across all of the specifications.

                                         [Insert Table 4 here]

       We next examine return dynamics around the announcement of active short selling
campaigns. Figure 2 plots CARs for the [t–100, t+100] period using the three-factor
model. Results are qualitatively similar with other models. The black line in Panel A
plots the CARs associated with active short selling campaigns. Return dynamics differ
dramatically in the periods before and after the announcement of a campaign. Abnormal
returns are positive prior to a campaign, reaching approximately 4%. The positive returns
suggest campaigns may partially be a response to increases in targets’ valuations. This
may particularly be the case for short sellers with existing positions in the target, who may
be under increased pressure due to rising share prices. Following announcements, CARs
drop to -7%, eventually stabilizing around -4%.10 The economic magnitude of this effect is
similar to abnormal returns associated with litigation related to financial fraud (Fich and
Shivdasani, 2007).

   9
     The positive CARs for the [t–10, t+10] period result from positive drift in the [t–10, t-1] period (as
seen in Figure 2).
  10
     Untabulated results indicate no reversal in abnormal returns when we consider [-100, 200] period
around the announcement of short selling campaign by hedge funds.

                                                    10
[Insert Figure 2 here]

     The dashed line in Panel A of Figure 2 plots CARs associated with large changes
in short interest. CARs prior to such events are virtually indistinguishable from those
for active short selling campaigns. While CARs decrease following large changes in short
interest, the economic magnitude of this effect is considerably smaller than for active short
selling campaigns.
     Panel A also provides insights into abnormal turnover around active short selling
campaigns. The dark bars in Figure 2 correspond to abnormal daily turnover. Abnormal
turnover is negative during the [t–100, t–20] period and positive during [t–10, t+30] period.
Abnormal turnover jumps after announcements and remains elevated for approximately one
month before decreasing to close to zero in the [t+30, t+100] period.
     Panel B of Figure 2 separately analyzes CARs for campaigns announced at investment
conferences and those announced in other ways (e.g., through the media). One advantage of
the conference sample is that the timing of announcements is likely orthogonal to other firm-
specific events that may confound the analysis. The light gray and black lines show CARs
for campaigns announced at conferences and the rest of the sample, respectively. Prior to
announcements, CARs for the conference sample are higher than other campaigns. After
the announcement, CARs for both groups fall by a similar magnitude. Thus, our findings
suggest that recent stock performance is an important factor for campaigns announced at
conferences, perhaps because hedge funds disclose stronger cases of overvalued securities.
Nevertheless, the evidence indicates confounding events do not drive the negative relation
between returns and active short selling.
     Overall, consistent with other work that studies the voluntary disclosure of short
positions in other contexts (e.g., Ljungqvist and Qian, 2016; Zhao, 2019), active short
selling campaigns by hedge funds are associated with strongly negative abnormal returns.
This finding validates our empirical setting by indicating market participants are largely
unaware of campaigns prior to public disclosure. In the next section, we explore whether
the effect of active short selling on shareholder wealth partially results from changes in

                                             11
behavior by other stakeholders that may negatively affect firms.

4. Stakeholder Behavior

     Active short selling potentially entails significant costs. In equilibrium, hedge funds
will undertake campaigns only if their expected benefits outweigh these costs. In the
previous section, we reported evidence of one such benefit: the flow of negative information
into prices as captured by CARs. Next, we examine whether campaigns affect the behavior
of other stakeholders in ways that may be detrimental to firms. Anecdotal evidence suggests
other stakeholders play important roles in active short selling campaigns. For example,
while refuting allegations made by David Einhorn, the management of Allied Capital argued
that tactics used by short sellers include creating “the illusion of a groundswell of concern”
by coordinating with other stakeholders (Einhorn, 2010). Our analysis focuses on short
selling by other market participants, media coverage, and litigation by shareholders and
other parties. We hypothesize that active short selling campaigns lead to changes in the
behavior along these dimensions that may have a negative effect on firm value.

4.1. Short Interest

     A potential benefit of active short selling (from the perspective of hedge fund) is that
it may induce a “bear raid” in which other investors sell the stock (e.g., Goldstein et al.,
2013; Khanna and Mathews, 2012). In this section, we test whether there are changes in
aggregate short selling behavior around the announcement of campaigns. To address this
question, we estimate the following regression:

                                 yit = αT argetit + ζi + εit ,                            (1)

where yit is aggregate short interest (normalized by shares outstanding) for event i and
period t, where a period is two weeks (due to the bi-weekly reporting of short interest).
T argetit is an indicator equal to one following the announcement of an active short selling

                                              12
campaign, and ζi are event fixed effects.11 The sample covers observation from 10 bi-weekly
periods before the announcement of a campaign to 10 periods after. The results of this
analysis are reported in Table 5.

                                           [Insert Table 5 here]

       The positive coefficient for the target indicator (both including and excluding event
fixed effects in columns 1 and 2, respectively) indicates that total short interest increases
after the announcement of a campaign. The economic magnitude of the effect is sizable,
corresponding to over a 10% increase in short interest for the campaigns in our sample. The
coefficient is significant at the 1% level when we control for time-invariant heterogeneity
at the event level. Columns 3 and 4 report similar results when we consider campaigns
announced at investment conferences.
       Panel A of Figure 3 plots total short interest for the 20 two-week periods around
the announcement of an active short selling campaign. There is an upward trend in short
interest prior to the announcement, potentially a consequence of hedge funds accumulating
their positions. Following the announcement, average short interest continues to increase
for approximately 5 periods (i.e., 10 weeks) and then remains relatively stable. The dashed
line in Panel A plots short interest for the sample of campaigns revealed at investment
conferences. The pre-trend is less apparent for this sample; short interest only increases
following the announcement.

                                          [Insert Figure 3 here]

       As an alternative measure of short selling activity, we examine fails-to-deliver (FTD).
A FTD may indicate that speculators are engaging in naked short selling (i.e., when a stock
is not borrowed prior to the settlement day), potentially resulting from an increase in short
selling activity (demand for borrowing shares) or a drop in the number of shares available

  11
    In contrast to later tests, we do not include firm-level accounting controls in the regression specification
due to the relatively high-frequency nature of short interest data.

                                                      13
to borrow (supply of shares available to borrow). Critics argue that naked short selling can
be “abusive” and exacerbate drops in prices for heavily shorted securities, though some
studies dispute this claim (Fotak et al., 2014).
      Panel B of Figure 3 plots the percentage fails-to-deliver from 100 days before the
announcement of a campaign to 100 days after.           The solid line plots the percentage
fails-to-deliver for the full sample and the dashed dark line plots the percentage fails-
to-deliver for short selling campaigns revealed at investment conference. The percentage of
FTD increases sharply during the ten day period prior to campaign announcement. The
percentage of FTD increases from about 0.04% to about 0.08% on the announcement date.
In unreported analysis, we find that the increase in FTDs in the [t–10, t+10] period is
statistically significant at conventional levels.
      Overall, the evidence suggests that active short selling campaigns are associated with
changes in trading behavior by other market participants. To the extent that an increase
in short selling leads to price declines, this effect may benefit hedge funds undertaking
campaigns. Moreover, the findings shed light on factors that influence the timing of
announcements. The sharp increase in FTDs for the full sample suggest that a decline in
the number of shares available to borrow might be a factor that triggers the announcement
of an active short selling campaign. However, when we consider campaigns announced
at investor conferences, we find no changes in the percentage of FTD around campaign
announcements, suggesting the strategic motives for the disclosure of active short selling
campaigns at investment conferences differ from those disclosed through other channels.

4.2. Media Coverage

      We next turn attention to the relation between short selling campaigns by hedge
funds and media coverage. Previous work finds that media coverage mitigates informational
frictions (e.g., Fang and Peress, 2009; Engelberg and Parsons, 2011), and the tone of such
coverage affects returns and volume (e.g., Tetlock, 2007). One potential benefit of public
disclosure of short positions is that it may influence the likelihood or tone of media coverage
of a firm. Such changes may, in turn, further reduce prices and therefore increase the value

                                               14
of funds’ short positions.
      To study media coverage around short selling campaigns, we estimate the following
OLS regression at the quarterly frequency:

                             yit = αT argetit + Xit0 β + ζt + ζi + εit ,                     (2)

where yit is outcome variable related to media coverage for firm i and year-quarter t,
T argetit is an indicator equal to one for one year (4 quarters) if a firm is targeted by an
active short selling campaign. Xit is a vector of control variables, ζt are year-quarter fixed
effects, and ζi are firm fixed effects.
      Table 6 reports results for the tone of media coverage. In this table, we report
estimates of regression (2), while considering only strongly negative articles (columns 1
through 4) and strongly positive articles (columns 5 through 8). The sentiment of media
coverage as measured by the Event Sentiment Score (ESS) assigned by Ravenpack. We
define an article as strongly negative if ESS70.

                                      [Insert Table 6 here]

      Columns 1–4 indicate that active short selling campaigns are associated with an
increase in the number of strongly negative articles.            The estimates are statistically
significant at the 1% level when we control for short interest (column 3) and other firm
characteristics (column 4), suggesting changes in media coverage are not explained by
aggregate short interest.     Similarly, columns 5–8 indicate a corresponding increase in
strongly positive articles. One potential explanation for this findings is that firms initiate
positive coverage to counter short sellers.
      Interpreting the relative economic magnitudes of the effects on positive and negative
articles is challenging because we use the natural logarithm of one plus the dependent
variable (due to some firms having zero positive or negative articles). To facilitate the
interpretation of the magnitudes, Figure 4 plots the percentage increase relative to the
sample mean for each ESS bin. The largest changes in the tone of media coverage are

                                                 15
driven by negative articles. While there is a statistically significant increase in coverage for
8 of the 10 bins, the three largest increases—ranging from 90% to 110%—are observed for
the most negative ESS bins.

                                         [Insert Figure 4 here]

       Overall, active short selling campaigns by hedge funds are associated with an increase
in media coverage, particularly for articles with a negative tone. This finding is consistent
with the idea that hedge funds may use the media as a platform to disseminate negative
information about targets. Indeed, the majority of campaigns (57.1%) in our sample are
initially announced via TV or newspaper. Thus, changes in the behavior of the media may
play an important role in active short selling campaigns.

4.3. Lawsuits

       Finally, we consider whether active short selling campaigns are associated with
changes in the likelihood of litigation. The disclosure of negative information about firms
may lead to lawsuits by regulators, shareholders, or other parties harmed. Such lawsuits
may lower firm value due to costs associated with mounting a defense or as part of a
settlement or penalty. Indeed, according to one survey of Fortune 200 firms, total litigation
costs averaged over $100 million per firm in 2008.12 Moreover, litigation may also be costly
for targeted firms if it leads to a change in corporate behavior (e.g., discontinuing fraudulent
practices).
       The results of our analysis are reported in Table 7. The outcome variables are an
indicator for litigation in the year following a short selling campaign (columns 1–4) and
the natural log of one plus the total number of cases in the year following a short selling
campaign (columns 5–8).13 The table reports estimates of the following OLS regression at

  12
     See       http://www.uscourts.gov/sites/default/files/litigation_cost_survey_of_major_
companies_0.pdf.
  13
     The types of lawsuits considered include shareholder litigation, fraud or accounting suits (potentially
initiated by the SEC or other regulators), IP suits, product liability suits, and antitrust suits.

                                                    16
the annual frequency:
                            yit = αT argetit + Xit0 β + ζt + ζi + εit ,                      (3)

where yit is the litigation-related outcome for firm i in year t, T argetit is an indicator equal
to one in the year following the announcement of an active short selling campaign , Xit is
a vector of control variables, ζt are year fixed effects, and ζi are firm fixed effects.

                                     [Insert Table 7 here]

     Our findings indicate active short selling is associated with an increase in litigation.
Specifically, columns 2–4 report an increase in the likelihood of litigation (within firm)
of approximately 11–13 percentage points. This effect is economically large relative to
the sample mean of 11%. We find similar results for the number of lawsuits faced by
firms in columns 5–8. In untabulated analysis, we consider specific types of lawsuits and
find a positive and statistically significant effect for shareholder lawsuits, fraud/accounting
lawsuits, and IP lawsuits. We do not, however, find evidence of a change in antitrust
lawsuits or product liability lawsuits. Importantly, the findings are robust to the inclusion
of controls for aggregate short interest (columns 3–4 and 7–8). Thus, while previous research
indicates short interest is associated with litigation or regulatory enforcement (e.g., Karpoff
and Lou, 2010), this effect does not drive our findings.
     In sum, our findings indicate that active short selling campaigns are associated with
a significant increase in litigation against targeted firms. These results are consistent with
the idea that campaigns reveal damaging information (e.g., fraud) that may be costly for
firms.

5. Discussion of Economic Mechanisms

5.1. Fund Characteristics

     To shed light on the question of why hedge funds engage in active short selling, we
first analyze the characteristics of investors that undertake campaigns. We estimate the

                                                17
following OLS regression at the fund-quarter level:

                                                  0
                              Campaignf t = Xf t α2 + ζt + ζf + εf t ,                              (4)

where the dependent variable Campaignf t is a dummy variable equal to one if fund
f initiates a short selling campaign during year-quarter t, Xf t is a vector of fund
characteristics from Schedule 13F filings, ζt are year-quarter fixed effects, and ζf are fund
fixed effects. Table A2 reports descriptive statistics for the fund-level variables used in
this analysis. Regressions cover Schedule 13F fund-quarter observations for all hedge funds
from 1999 through 2012.14 Table 8 reports the results.

                                        [Insert Table 8 here]

       Columns 1 and 2 of Table 8 use the full sample of active short selling campaigns (both
with and without fund fixed effects). Column 1 indicates the likelihood of undertaking
a campaign is 0.56% higher for activist hedge funds than for non-activist hedge funds.
The economic magnitude of this coefficient is approximately three times the unconditional
likelihood of an active short selling campaign. This finding is consistent with the fact that
activists initiate 208 out of 252 campaigns in our sample (82%). Importantly, the effect
is not driven by a higher representation of activists in the sample of hedge funds; in our
dataset, 240 out of 1,199 hedge funds are activists.
       Other fund characteristics have less explanatory power. The likelihood of a campaign
is positively correlated with assets under management; in terms of economic magnitudes, a
10% increase in assets under management is associated with approximately a 5% increase
in the likelihood of an active short selling campaign. We also find some evidence that funds
undertaking campaigns tend to have smaller long positions in heavily shorted securities,
though this effect is not robust across all specifications. We do not find evidence that
other observable characteristics of funds (e.g., returns, portfolio turnover rate, portfolio

  14
    We end the sample for this test in 2012 due to coverage of the data on options use. In untabulated
results, we find qualitatively similar results for the other fund characteristics using the full sample.

                                                  18
illiquidity, number of holdings, use of put options, etc.) are associated with active short
selling. We obtain similar results for the sample of campaigns undertaken by activist hedge
funds (columns 3 and 4 in Table 8).
     In sum, our findings indicate that shareholder activism experience plays a critical
role in understanding active short selling. In the remainder of this section, we examine
economic factors that potentially explain this behavior.

5.2. Active Short Selling and Activism Technology

     The dominance of activist hedge funds in our sample (82% of campaigns) suggests
that the distinction between activists and non-activists is important for understanding
active short selling. We next consider economic factors that may explain this difference in
behavior.
     Our analysis focuses on three potential channels: benefits of active short selling,
costs of active short selling, and the information sets of investors.      First, campaigns
are associated with benefits including negative announcement returns and changes in
stakeholder behavior. One possibility is that differences between activists and non-activists
lead to heterogeneity in these benefits. For example, activists may have more skill in
obtaining media exposure, thus facilitating campaigns. Second, we consider the costs
associated with active short selling. It is possible that these costs differ for activist and
non-activist hedge funds. For example, activists often engage in hostile activism campaigns,
so the additional costs of public confrontation may be relatively low. Finally, differences
in the information sets of activist and non-activist hedge funds may explain this behavior.
Specifically, the ability to engage in activism incentivizes investors to uncover problems at
firms. If such problems are not priced, this creates an incentive to undertake an active
short selling campaign.

5.2.1. Campaign Benefits
     Our previous findings highlight potential benefits associated with active short selling.
We begin our analysis by asking whether these benefits differ for activist and non-activist

                                             19
hedge funds.    One caveat of this analysis, however, is that activists undertake most
campaigns in our sample. Consequently, the statistical power of our tests is limited.
     Appendix Table A3 reports the difference in announcement CARs for campaigns
initiated by activists and non-activists. Overall, the estimates are economically small
and statistically indistinguishable from zero. In untabluated analysis, we also find little
evidence of differences in the behavior of other stakeholders (short selling, media coverage,
and litigation) for campaigns undertaken by activists versus non-activists. Overall, these
findings suggest that differences in the expected benefits of active short selling are not likely
to explain the relation between activism technology and active short selling campaigns.

5.2.2. Campaign Costs
     We next consider the possibility that the costs associate with active short selling
differ for activist and non-activist hedge funds. It is important to separately consider
two distinct types of costs related to campaigns. First, there are costs associated with
retaining short positions. If short sale constraints are more binding for activists than
for non-activists, this may serve as an explanation for why activists engage in active short
selling. Consistent with this idea, Ljungqvist and Qian (2016) show that small, constrained
investors release research reports to mitigate limits to arbitrage. However, it is unlikely
that such costs explain our findings. For one, while both activists and non-activists incur
costs associated with retaining short positions, campaigns are predominately undertake by
activists. Indeed, activists in our sample average 0.59 campaigns each, while non-activists
average 0.03. Moreover, Table 8 shows that the likelihood of undertaking a campaign is
positively related to assets under management, the opposite effect predicted by the limits
to arbitrage channel.
     Active short selling is also potentially costly due to the scrutiny, investigations, and
litigation that may accompany campaigns. Because activism campaigns often involve the
use of hostile tactics to influence corporate policies (Brav et al., 2008), activists may
be uniquely suited to bear such costs. That is, because many activists already have a
reputation for being hostile, the incremental costs associated with further confrontations

                                               20
may be relatively low.
       To test this idea, we examine the characteristics of activists that undertake active
short selling campaigns. We first examine the relationship between previous experience
with long shareholder activism and active short selling. We hypothesize that the costs
associated with active short selling are lower for investors that have previously engaged
in more activism campaigns. Panel A of Table 9 sorts activist hedge funds into quartiles
based on their number of previous long activism campaigns. There is a wide range of
experience among the activist hedge funds in our sample; those in the first quartile have
undertaken just one activism campaign, while those in the fourth have undertaken 22.6
activism campaigns on average.

                                          [Insert Table 9 here]

       For each quartile, we report the average number of active short selling campaigns in
column 4. Activists in quartiles 1 and 2 (i.e., those with the least activism experience)
initiated 0.37 and 0.32 campaigns, respectively. However, those in quartiles 3 and 4 (with
the most activism experience) initiated 1.15 and 0.76 campaigns, respectively. Thus, the
evidence indicates that experience with long shareholder activism is positively correlated
to active short selling.
       Panel B examines whether there are differences in the tactics used for long activism
campaigns by hedge funds with and without active short selling experience. Activist hedge
funds that undertake active short selling campaigns tend to be more hostile in long activism
campaigns than those that do not.15 Specifically, the likelihood of initial hostility is 4.8
percentage points higher for activist hedge funds with short selling campaign experience,
while the likelihood of hostility at any point during activism campaign is 4.4 percentage
points higher. In addition, there is evidence that funds that engage in active short selling
have more ambitious goals in activism campaigns; the likelihood of seeking a sale is 5.7

  15
     Brav et al. (2008) define hostile campaigns that includes actions such as threatening/initiating a proxy
fight, suing the company, or intending to take control of the company.

                                                     21
percentage points higher and the likelihood of seeking a change in business strategy is 4.4
percentage points higher for activist hedge funds with active short selling experience.
     Overall, the evidence suggests costs associated with active short selling are an
important factor for understanding differences in the behavior between activist and non-
activist hedge funds. Specifically, active short selling campaigns are undertaken by hedge
funds with more activism experience as well as those that employ hostile tactics and have
more ambitious (and potentially contentious) goals. To the extent that these characteristics
limit the costly nature of public scrutiny, the findings suggest a reason why access to
activism technology facilitates active short selling.

5.2.3. Information
     Finally, we consider the role of information in explaining the sharp difference in active
short selling engagement for activist and non-activist hedge funds. The investment strategy
of activist investors focuses on identifying problems with firms. This is a fundamental
difference between activists and other investors. David Einhorn, an activist who undertakes
26 campaigns in our sample, highlights this difference:

      “A typical process to identify opportunities is through computer screens that
      identify statistical cheapness, such as low multiples of earnings, sales, or book
      value combined with rising earnings estimates. Then, they evaluate the identified
      companies as possible investments...Greenlight takes the opposite approach. We
      start by asking why a security is likely to be misvalued in the market. Once we
      have a theory, we analyze the security to determine if it is, in fact, cheap or
      overvalued ” (Einhorn, 2010).

     Once an activist identifies a problem at a target, they must then determine if (1) the
problem is priced and (2) the problem can be corrected. If the problem is priced and can
be fixed via an intervention, the activist may initiate an activism campaign. The literature
on hedge fund activism has studied these campaigns.
     Our findings indicate that when a problem is not priced, activists may initiate an

                                              22
active short selling campaign. For example, activists often undertake a campaign when
there is a problem related to a target’s business model (see Table 1). The fact that activists
engage in active short selling after identifying this problem indicates that they believe the
problem is not priced. For some accusations (e.g., fraud) the economic incentives for long
shareholder activism are absent because revealing private information will negatively affect
the target’s value. However, there are strong economic incentives to establish a short
position and engage in a short selling campaign in such instances.
     We conjecture that activism technology may facilitate active short selling because
activists have more information on firms’ specific problems. Of course, testing for differences
in information sets is an inherently difficult task. To shed some light on this economic
mechanism, we undertake further analysis of CARs associated with active short selling
campaigns. Specifically, we investigate whether stock price reactions depend on allegations
of short selling campaigns.     We classify allegations into two broad samples: general
overvaluation and specific allegations. Specific allegations—financials/capital structure, in-
dustry/competitors, fraud/accounting, product/business model, and management/insider
trading—point to a particular problem with the target.
     The dark line in Figure 5 corresponds to CARs for campaigns with specific allegations.
Prior to announcement, abnormal returns are close to zero. On the announcement date,
target stocks experience sharp negative abnormal returns, reaching -10% two weeks after
the announcement. Panel A of Table 10 shows that the negative abnormal returns are
statistically significant and differ from CARs associated with large changes in short interest.

                                    [Insert Figure 5 here]

     The grey line in Figure 5 corresponds to CARs for campaigns with general
overvaluation allegations. Price dynamics for these campaigns are strikingly different.
Targets of these campaigns experience large positive abnormal returns prior to campaign
announcement.      Specifically, abnormal returns reach 10% during [t-100,t-1] period,
consistent with the allegation that targets are overvalued.        During two months after
campaign announcement, targets experience abnormal return of about -5%. Panel B of

                                              23
Table 10, however, shows that these abnormal returns are statistically insignificant and are
not statistically different from CARs for large increases in short interest.
     Our findings highlight stark differences in CARs based on the nature of allegations
from active short selling campaigns. When allegations pertain to general overvaluation,
stock price reactions are similar to those for large increases in short interest. In contrast, if
allegations pertain to a specific problem with the target, CARs are significantly negative.
Thus, our findings are consistent with the idea that differences in the information sets of
activists and non-activists may explain, at least in part, why activism technology facilitates
active short selling.   Specifically, because the investment strategy of activists focuses
on identifying problems at firms, activists are more likely uncover private information
associated with large price reactions.
     In sum, the evidence suggests two factors play an important role in explaining why
access to activism technology plays an important role in active short selling campaigns.
First, costs of public confrontation with targets may be lower for activist hedge funds than
for non-activist hedge funds. Second, because the investment strategy of activists focuses
on identifying specific problems (e.g., rather than using quantitative signals), they are more
likely to find information that is conducive to active short selling.

5.3. Alternative Mechanisms

     Next, we consider alternative mechanisms that may explain the findings.                 We
specifically consider reputational effects and hedging motivations for active short selling.
We do not find evidence consistent with these alternative explanations.

5.3.1. Reputation Effects
     We first consider if active short selling is associated with reputational effects.
Specifically, successful campaigns may improve the reputation of activist hedge funds and
lead to higher returns for subsequent long activism campaigns. This may be the case,
for example, if successful campaigns lead to more support in subsequent campaigns from
long-term investors (Appel et al., 2018). To test this idea, we regress long campaign CARs

                                               24
You can also read